benchgap
Agentic · tools

OSWorld-Verified leaderboard

As of 2026-11-21, the highest measured score on OSWorld-Verified is 86.1% by Qwen3.8 Max. 172 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Opus 5.591.6%estimated ± 8.4 pp, low confidence
2Muse Wonder Full91.3%estimated ± 8.4 pp, low confidence
3Claude Sonnet 5.590.8%estimated ± 8.4 pp, low confidence
4Muse Wonder89.0%estimated ± 8.4 pp, low confidence
5Claude Fable 5.188.6%estimated ± 8.4 pp, low confidence
6Grok 4.787.5%estimated ± 8.4 pp, low confidence
7Kimi K3.1 Pro87.1%estimated ± 5.7 pp, low confidence
8GPT-6 Astra86.4%estimated ± 5.7 pp, low confidence
9Qwen3.8 Max Preview86.3%estimated ± 8.4 pp, low confidence
10Qwen3.8 Max86.1%measured
11Claude Opus 585.7%estimated ± 2.3 pp, low confidence
12Gemini 4 Argon85.2%estimated ± 8.4 pp, low confidence
13Claude Haiku 5.585.1%estimated ± 8.4 pp, low confidence
14Claude Fable 585.0%measured
15Claude Mythos 585.0%measured
16Inkling 2 Small84.9%estimated ± 5.7 pp, low confidence
17GPT-5.5 Pro84.8%estimated ± 5.7 pp, low confidence
18GPT-6.1 Sol84.4%estimated ± 8.4 pp, low confidence
19Qwen3.8-27B84.3%measured
20GLM-5.3-Flash83.9%estimated ± 9.0 pp, low confidence
21GPT-5.4 Pro83.9%estimated ± 5.7 pp, low confidence
22Ultramarine 1 Flash83.5%estimated ± 8.4 pp, low confidence
23Claude Opus 4.883.4%measured
24Gemini 3.6 Flash83.0%measured
25Qwen3.8-Flash-Next82.8%estimated ± 3.0 pp, high confidence
26Holo3-35B-A3B82.6%measured
27GPT-6 Sol82.1%estimated ± 8.4 pp, low confidence
28GPT-5.6 Sol82.0%estimated ± 2.3 pp, medium confidence
29MiMo-V2.6-Pro82.0%measured
30ERNIE 5.5 Flash Preview81.6%estimated ± 5.7 pp, medium confidence
31Gemini 3.8 Flash81.3%estimated ± 2.3 pp, medium confidence
32Claude Sonnet 581.2%measured
33Muse Spark 1.281.1%estimated ± 8.4 pp, low confidence
34MiMo-V2.6-Flash80.8%measured
35Muse Spark 1.180.8%measured
36DeepSeek V4.1 Flash 101780.7%estimated ± 4.7 pp, high confidence
37Ling 3.1 Flash80.7%estimated ± 4.7 pp, high confidence
38Fugu Cyber80.5%estimated ± 4.7 pp, high confidence
39Atria Dawn Preview80.4%estimated ± 4.7 pp, high confidence
40Gemini 3.8 Flash Cyber80.3%estimated ± 4.7 pp, high confidence
41GPT-6 Luna80.1%estimated ± 8.4 pp, low confidence
42Step 5 Preview80.0%estimated ± 4.7 pp, high confidence
43GLM-5.380.0%estimated ± 4.7 pp, high confidence
44DeepSeek V4 Pro79.7%estimated ± 4.7 pp, high confidence
45Gemini 3.5 Flash Cyber79.7%estimated ± 4.7 pp, high confidence
46Claude Mythos Preview79.7%estimated ± 4.7 pp, high confidence
47Muse Spark 1.379.5%estimated ± 2.3 pp, medium confidence
48MiniMax M3.1 Flash79.5%estimated ± 4.7 pp, high confidence
49Muse Spark Open79.3%estimated ± 8.4 pp, low confidence
50Kimi K379.1%estimated ± 2.3 pp, medium confidence
51Apodex 1.178.9%estimated ± 7.1 pp, low confidence
52Holo3-122B-A10B78.9%measured
53Grok 4.578.8%estimated ± 8.4 pp, low confidence
54GPT-5.578.7%measured
55Hy4 preview78.6%estimated ± 4.7 pp, high confidence
56Gemini 3.7 Flash78.4%estimated ± 2.3 pp, medium confidence
57GPT-5.6 Terra78.4%estimated ± 2.3 pp, medium confidence
58Gemini 3.5 Flash78.4%measured
59DeepSeek V4 Flash 073178.2%estimated ± 4.7 pp, high confidence
60Mistral Large 478.1%estimated ± 5.7 pp, medium confidence
61Claude Opus 4.778.0%measured
62GLM-5.277.7%estimated ± 5.1 pp, medium confidence
63Grok 4.677.3%estimated ± 2.3 pp, medium confidence
64UI-Mate-27B77.0%measured
65GPT-5.6 Luna75.1%estimated ± 2.3 pp, medium confidence
66GPT-5.475.0%measured
67Sage-A1.574.7%estimated ± 5.7 pp, medium confidence
68Ornith-1.5-397B74.2%estimated ± 3.8 pp, low confidence
69Gemini 3.5 Flash-Lite74.0%measured
70LongCat-2.573.9%estimated ± 3.8 pp, low confidence
71K-EXAONE 2.073.9%estimated ± 3.8 pp, low confidence
72Ornith-1.0-397B73.8%estimated ± 3.8 pp, low confidence
73Apodex 1.1 Mini73.6%estimated ± 8.4 pp, low confidence
74Gemini 3 Pro73.6%estimated ± 5.5 pp, medium confidence
75Quasar 438B73.5%estimated ± 8.4 pp, low confidence
76dots3-note Preview73.3%estimated ± 3.8 pp, medium confidence
77Qwen3.7 Plus73.3%measured
78Ornith-1.5-35B-A3B73.2%estimated ± 3.8 pp, medium confidence
79Qwen3.6-27B73.2%estimated ± 3.8 pp, medium confidence
80Kimi K2.673.1%measured
81Ling 3.0 Flash Sante72.9%estimated ± 5.7 pp, medium confidence
82Ling 3.0 Flash VL72.9%estimated ± 8.4 pp, low confidence
83Spark-X2.6-4B72.7%estimated ± 3.8 pp, medium confidence
84Claude Opus 4.672.7%measured
85Ornith-1.0-35B72.7%estimated ± 3.8 pp, medium confidence
86Qwen3.6-35B-A3B72.4%estimated ± 3.8 pp, medium confidence
87LFM3-1.7B72.4%estimated ± 3.8 pp, medium confidence
88Claude Sonnet 4.672.1%measured
89GPT-5.4 mini72.1%measured
90Step 3.7 Flash72.0%estimated ± 3.8 pp, medium confidence
91Ornith-1.5-9B71.9%estimated ± 3.8 pp, medium confidence
92Inkling 2 Mini71.5%estimated ± 5.7 pp, medium confidence
93Ripple-65B Preview71.5%estimated ± 9.4 pp, low confidence
94Qwen3.7 Max71.5%estimated ± 3.8 pp, medium confidence
95LLaDA2.2-flash71.1%estimated ± 3.8 pp, medium confidence
96MiMo-V2.5-Pro71.0%estimated ± 3.8 pp, medium confidence
97Muse Spark71.0%estimated ± 3.8 pp, medium confidence
98Ornith-1.0-9B70.7%estimated ± 3.8 pp, medium confidence
99LFM2.5-2.6B70.6%estimated ± 3.8 pp, medium confidence
100GLM-5.170.4%estimated ± 3.8 pp, medium confidence
101MiMo-V2.570.4%estimated ± 3.8 pp, medium confidence
102Nemotron 3.5 Super VL 120B A12B70.4%estimated ± 5.7 pp, medium confidence
103Hy370.3%estimated ± 8.4 pp, low confidence
104MiniMax M370.1%measured
105Nemotron 3.5 Super VL 120B A12B NVFP469.8%estimated ± 5.7 pp, medium confidence
106Inkling-Small69.7%estimated ± 5.7 pp, medium confidence
107Kimi K2.7 Code69.6%estimated ± 8.4 pp, low confidence
108Inkling69.4%estimated ± 5.7 pp, medium confidence
109Beam 1.0 501B-A23B69.0%estimated ± 5.7 pp, medium confidence
110Beam 1.0 501B-A23B FP868.9%estimated ± 5.7 pp, medium confidence
111Qwen3.6 Plus68.6%estimated ± 3.8 pp, medium confidence
112Mercury 368.3%estimated ± 7.1 pp, low confidence
113Gemini 3.1 Pro67.9%estimated ± 3.8 pp, medium confidence
114MiMo-V2-Pro67.9%estimated ± 3.8 pp, medium confidence
115GLM-567.9%estimated ± 3.8 pp, medium confidence
116Beam 1.0 501B-A23B NVFP467.7%estimated ± 5.7 pp, medium confidence
117Agents-A167.6%estimated ± 5.7 pp, medium confidence
118A.X K267.6%estimated ± 8.4 pp, low confidence
119LLaDA2.2-mini67.5%estimated ± 3.8 pp, medium confidence
120GPT-566.5%estimated ± 8.4 pp, low confidence
121GLM-5-Turbo66.4%estimated ± 3.8 pp, medium confidence
122Claude Opus 4.566.3%measured
123UI-Mate-9B66.2%measured
124Muse Glimmer 30B65.9%measured
125Solar Open 264.9%estimated ± 9.4 pp, low confidence
126GPT-5.3 Codex64.7%measured
127GLM-5V-Turbo64.6%estimated ± 3.8 pp, medium confidence
128Ling 3.0 Flash64.2%estimated ± 5.7 pp, medium confidence
129Kimi K2.563.3%measured
130GPT-5.2-Codex63.1%estimated ± 5.5 pp, medium confidence
131Qwen3.5 397B62.2%measured
132Nemotron 3.5 Bolt 74B-A9B BF1662.1%estimated ± 5.7 pp, medium confidence
133Grok 4.161.6%estimated ± 5.1 pp, low confidence
134Claude 4.1 Opus61.5%estimated ± 8.4 pp, low confidence
135Claude Sonnet 4.561.4%measured
136GPT-5.1-Codex61.4%estimated ± 5.5 pp, medium confidence
137Qwen3.5 Plus61.0%estimated ± 8.4 pp, low confidence
138Grok Build 0.161.0%estimated ± 5.5 pp, medium confidence
139MiniCPM5-2B60.9%estimated ± 8.4 pp, low confidence
140Claude Haiku 4.560.8%estimated ± 8.4 pp, low confidence
141Grok 4.1 Fast59.7%estimated ± 5.5 pp, medium confidence
142Agents-A1-4B59.4%estimated ± 5.7 pp, medium confidence
143Nemotron 3.5 Bolt 74B-A9B59.4%estimated ± 5.7 pp, medium confidence
144MiMo-V2-Flash58.8%estimated ± 8.4 pp, low confidence
145Gemini 3 Flash58.7%estimated ± 3.8 pp, low confidence
146Qwen3.5-122B-A10B58.0%measured
147MiniMax M2.757.9%estimated ± 3.8 pp, low confidence
148Qwen3 Max57.5%estimated ± 5.5 pp, medium confidence
149Granite 4.2 30B57.4%estimated ± 8.4 pp, low confidence
150Gemma 4 26B A4B57.1%estimated ± 8.4 pp, low confidence
151Ling 3.0 Tiny57.0%estimated ± 8.4 pp, low confidence
152Grok 456.8%estimated ± 5.5 pp, medium confidence
153Gemini 2.5 Pro56.6%estimated ± 5.5 pp, medium confidence
154GPT-5.156.3%estimated ± 5.5 pp, medium confidence
155Qwen3.5-27B56.2%measured
156GLM-4.755.8%estimated ± 5.5 pp, medium confidence
157Command A+55.8%estimated ± 8.4 pp, low confidence
158Claude 4 Sonnet55.7%estimated ± 5.5 pp, medium confidence
159Mistral Medium 3.5 128B55.5%estimated ± 5.5 pp, medium confidence
160Celeris-155.4%estimated ± 8.4 pp, low confidence
161Claude 3.5 Haiku55.4%estimated ± 8.4 pp, low confidence
162DeepSeek V355.4%estimated ± 8.4 pp, low confidence
163DeepSeek V3 032455.4%estimated ± 8.4 pp, low confidence
164Gemma 3 27B55.4%estimated ± 8.4 pp, low confidence
165Gemma 4 12B55.4%estimated ± 8.4 pp, low confidence
166Gemma 4 E2B55.4%estimated ± 8.4 pp, low confidence
167Gemma 4 E4B55.4%estimated ± 8.4 pp, low confidence
168GPT-4.1 mini55.4%estimated ± 8.4 pp, low confidence
169GPT-4.1 nano55.4%estimated ± 8.4 pp, low confidence
170GPT-4o55.4%estimated ± 8.4 pp, low confidence
171GPT-4o mini55.4%estimated ± 8.4 pp, low confidence
172GPT-OSS 20B55.4%estimated ± 8.4 pp, low confidence
173Granite 4.2 3B55.4%estimated ± 8.4 pp, low confidence
174Granite 4.2 8B55.4%estimated ± 8.4 pp, low confidence
175K-Exaone55.4%estimated ± 8.4 pp, low confidence
176Ling 2.6 Flash55.4%estimated ± 8.4 pp, low confidence
177Llama 3.3 70B55.4%estimated ± 8.4 pp, low confidence
178Llama 4 Maverick55.4%estimated ± 8.4 pp, low confidence
179Llama 4 Scout55.4%estimated ± 8.4 pp, low confidence
180Mercury 2.555.4%estimated ± 8.4 pp, low confidence
181Mistral Large 355.4%estimated ± 8.4 pp, low confidence
182Mistral Small 3.155.4%estimated ± 8.4 pp, low confidence
183Mistral Small 455.4%estimated ± 8.4 pp, low confidence
184Nemotron 3 Nano 30B55.4%estimated ± 8.4 pp, low confidence
185Nemotron 3 Nano Omni 30B A3B55.4%estimated ± 8.4 pp, low confidence
186North Mini Code55.4%estimated ± 8.4 pp, low confidence
187Solar Pro 355.4%estimated ± 8.4 pp, low confidence
188Trinity-Large-Preview55.4%estimated ± 8.4 pp, low confidence
189Ultravox v0.6 Llama 3.3 70B55.4%estimated ± 8.4 pp, low confidence
190Gemini 3.1 Flash-Lite55.3%estimated ± 5.5 pp, medium confidence
191Grok 4.2055.3%estimated ± 5.5 pp, medium confidence
192Hy3 Preview54.8%estimated ± 5.5 pp, medium confidence
193Qwen3.5-35B-A3B54.5%measured
194Gemma 4 31B54.4%estimated ± 5.5 pp, medium confidence
195Trinity-Large-Thinking54.0%estimated ± 5.5 pp, medium confidence
196Grok 4.353.6%estimated ± 5.1 pp, low confidence
197GPT-OSS 120B53.6%estimated ± 5.5 pp, medium confidence
198GPT-4.153.3%estimated ± 5.5 pp, low confidence
199MiMo-V2-Omni51.2%estimated ± 3.8 pp, low confidence
200Sarvam 105B51.2%estimated ± 5.7 pp, low confidence
201Solar Pro 451.1%estimated ± 5.7 pp, low confidence
202LongCat-Flash-Lite-Sparse51.0%estimated ± 5.7 pp, low confidence
203Nemotron 3 Ultra50.3%estimated ± 5.7 pp, low confidence
204Nemotron 3.5 Lightning 30B A3B NVFP449.6%estimated ± 5.7 pp, low confidence
205Sarvam 30B49.6%estimated ± 5.7 pp, low confidence
206GPT-5.247.3%measured
207GPT-5.4 nano39.0%measured
208DeepSeek V3.238.7%estimated ± 3.8 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General